The Grouparoo Blog


Understanding Types with SQLite and Node.js

Tagged in Engineering Notes 
By Sean C Davis on 2021-04-22

SQLite Quirks

Two fun facts about SQLite:

  1. The initial release was more than 20 years ago!
  2. It is the most widely used database (and likely one of the most widely deployed pieces of software).

And here are a few of my opinions on SQLite:

  • It's super cool.
  • We don't talk about it enough.
  • It's actually really easy to use (which is likely why it's so widely used).
  • It is a little quirky.

So let's talk about this super cool thing. Let's look at how easy it is to use with Node.js before spending some time uncovering its quirks, which mostly have to do with how it handles data types.

A Quick SQLite Example with Node.js

If you're able to run Node on your machine and install packages via NPM, then you can very easily create and manipulate a SQLite database. Let's go through an example in which we do just that — create a database, users table, and a few users.

(Note: You can also work with SQLite without Node or NPM. That's just what we're using for our examples here today.)

To get started, create a new directory for your project:

$ mkdir my-sqlite-project
$ cd my-sqlite-project
$ npm init -y

Now that you have an empty directory, let's install our dependencies:

$ npm install sqlite3 faker

Next, create an index.js file with the following code:

const sqlite3 = require("sqlite3").verbose();
const faker = require("faker");
const path = require("path");
const { promisify } = require("util");

// Create a new database named mydb.sqlite in the root of this project.
const dbFilePath = path.join(__dirname, "mydb.sqlite");
const db = new sqlite3.Database(dbFilePath);

// Use the promise pattern for SQLite so we don't end up in callback hell.
const query = promisify(db.all).bind(db);

// SQL query for creating a users table if it doesn't already exist.
const createTableQuery = `
  CREATE TABLE IF NOT EXISTS users (
    "id" INTEGER PRIMARY KEY,
    "email" TEXT NOT NULL,
    "first_name" TEXT,
    "last_name" TEXT,
    "created_at" TEXT
  )
`;

// Generate user attributes using faker.
const newUser = {
  email: faker.internet.email(),
  first_name: faker.name.firstName(),
  last_name: faker.name.lastName(),
  created_at: Date.now(),
};

/**
 * Run an INSERT query on some given table and insert the given object.
 */
const create = async ({ table, object }) => {
  const keys = Object.keys(object).join(",");
  const values = Object.values(object)
    .map((v) => `"${v}"`)
    .join(",");
  const res = await query(`INSERT INTO ${table} (${keys}) VALUES (${values})`);
  return res;
};

/**
 * Read all records and all their columns from some given table.
 */
const read = async ({ table }) => {
  const res = await query(`SELECT * FROM ${table}`);
  return res;
};

/**
 * The main controller of this script. This is inside an async function so we
 * can use the promise pattern.
 */
const run = async () => {
  // Create users table if it doesn't exist.
  await query(createTableQuery);
  // Create a new user.
  await create({ table: "users", object: newUser });
  // Read all the users.
  const users = await read({ table: "users" });
  // Print to the console.
  console.log(users);
};

run();

Then run the script:

$ node index.js

After you do that, two things should happen:

  1. You should see some output in the console that is an array containing a single user with the values you just randomly generated. Something like:

    [
      {
        id: 1,
        email: "Dawson39@yahoo.com",
        first_name: "Dorris",
        last_name: "West",
        created_at: "1619034411275",
      },
    ];
    
  2. You should have a new file in your project called mydb.sqlite. This is your database!

You can open up your database file with a SQLite browser. I'm a big fan of TablePlus, which has a free version. But you can also use the simple-but-popular DB Browser for SQLite.

If you open your users table, you should see the same record represented as was printed to the console.

See how easy that was?

Now that we have that basic example in place, let's keep it around while we dig into some of SQLite's quirks.

SQLite Type Quirks

SQLite is weird. Well, it's not really weird, it's just ... simple. So it feels weird in comparison to other beefy (and structured) databases like PostgreSQL or MySQL.

Most of the quirks with SQLite have to do with how it handles types. And that's in large part because of how SQLite stores its data (as a file on your file system). That's right. That mydb.sqlite file you created in the simple example above is your entire database. Cool, right?

Let's dig in to the basics of SQLite's data types before we look at how we can better work with them when writing a program with Node.js.

Understanding SQLite Types

SQLite only has five data types:

  • NULL: The value is a NULL value.
  • INTEGER: The value is a signed integer, stored in 1, 2, 3, 4, 6, or 8 bytes depending on the magnitude of the value.
  • REAL: The value is a floating point value, stored as an 8-byte IEEE floating point number.
  • TEXT: The value is a text string, stored using the database encoding (UTF-8, UTF-16BE or UTF-16LE).
  • BLOB: The value is a blob of data, stored exactly as it was input.

What the heck does that mean? Reading between the lines, it means that the data in your database is either going to be (in JavaScript speak) a string or a number, at least for the majority of cases.

That's right. That means there are no dates or booleans in SQLite. WTF?

SQLite does not store date objects

SQLite has date functions, but it will ultimately store the actual value as either a string or number.

For example, I could run this query against the users table from above:

INSERT INTO users (email,created_at) VALUES ('hello@example.com',date('now'))

And the value will be stored as 2021-04-21, as a string.

SQLite also doesn't store booleans

Instead, booleans used in SQL statements are converted to either 0 or 1. In other words, true in an insert statement becomes 1 when it's stored in the database.

How Type Affinity Works in SQLite

SQLite uses what is called dynamic typing. Where most other databases set the data type of some given value by the column definition, SQLite looks more directly at the value itself.

But, there is some magic happening through what SQLite calls type affinity on columns in a database. You can set a type for a column, but it's really just a recommended type. SQLite will do its best to convert the value to that type, but it may not always work as you'd expect.

The goofy thing here is that aside from matching a type affinity name when setting the type of a column, the name is virtually meaningless. It can be anything. Here are a set of types that map to type affinities in SQLite.

Let's go through some specific examples to see what happens when we use the Node sqlite3 library to interact with a SQLite database.

Type Affinity Examples with SQLite and Node.js

Let's go through a few examples together to show how type affinity works with SQLite and how we can use JavaScript to gain more control over types by manually casting them. Then we'll close with a cool approach for getting around all this nonsense.

Using the example from the beginning of the post, let's add a few columns to our users table. Edit the createTableQuery in index.js:

const createTableQuery = `
  CREATE TABLE IF NOT EXISTS users (
    "id" INTEGER PRIMARY KEY,
    "email" TEXT NOT NULL,
    "first_name" TEXT,
    "last_name" TEXT,
    "created_at" TEXT,
    "a" TEXT,
    "b" INTEGER,
    "c" CHEESE
  )
`;

Here we've recreated the users table with three new columns:

  • a of type TEXT. This matches the TEXT affinity.
  • b of type INTEGER. This matches the INTEGER affinity.
  • c of type CHEESE. This has no affinity.

Then, let's set a, b, and c on our newUser object to be a random number.

const newUser = {
  email: faker.internet.email(),
  first_name: faker.name.firstName(),
  last_name: faker.name.lastName(),
  created_at: Date.now(),
  a: faker.datatype.number(),
  b: faker.datatype.number(),
  c: faker.datatype.number(),
};

Before we load the content into the database, let's check the type. And when we retrieve it from the database, we'll check the types again.

const run = async () => {
  await query(createTableQuery);

  const user = newUser();
  // Log the data types going into the database.
  Object.entries(user).map(([key, value]) =>
    console.log(`${key}: ${typeof value}`)
  );
  await create({ table: "users", object: user });

  const users = await read({ table: "users" });
  // Log the types coming out of the database.
  Object.entries(users[0]).map(([key, value]) =>
    console.log(`${key}: ${value} (${typeof value})`)
  );
};

Delete the database mydb.sqlite and run the script:

$ node index.js

Notice what you see on the console. Before the data went into the database, each of our new values was a number:

a: number
b: number
c: number

That's what we should have expected, considering we set those attributes as random numbers.

For me, when they came out of the database, they looked different:

a: '91057' (string)
b: 9807 (number)
c: 31711 (number)

Notice that a is now a string. Why? Because we set the column's type affinity to a value that SQLite will use to try to convert to a string. Meanwhile, nothing happened with the CHEESE column because it has no affinity, so it left it as a number.

Try again, but with strings

If we do that again, but change faker.datatype.number() to faker.datatype.string(), we will see something slightly different.

The types going in are all strings:

a: string
b: string
c: string

But coming out they are still all strings:

a: 'i_`kneb8|]' (string)
b: '/@adUCVEV3' (string)
c: '@8eMpbKoFk' (string)

The reason b wasn't converted to a number is because SQLite doesn't know how to convert that random string to a number. There's no logical choice for what the number should be. So it leaves the original value as is.

Using a number as a string

Now, one last time. Let's use a number again, but let's convert it to a string. Change faker.datatype.string() to faker.datatype.number.toString().

Now, going in we still have all strings:

a: string
b: string
c: string

But coming out, it looks a little different:

a: '42995' (string)
b: 22072 (number)
c: 3466 (number)

Whoa! Something weird happened this time. Now our affinities match for a and b as they did in the first example. This is example SQLite knows how to turn "22072" into an integer.

But c changed from being a string to a number. That's because SQLite is trying to do the best it can't with what it is given. And in this case it figured it could turn the string into a number, so it did that. The only way to have kept it as a string would have been to use an appropriate affinity mapping value.

Typecasting SQLite Queries using Node.js

This quirk is pretty interesting and clever, but it's dangerous in real-world scenarios when you don't have bounds around how it can behave. You don't want your database turning strings into numbers without knowing why that's happening.

One approach is to write your own getters and setters that typecast values more strictly. That way your program has more control over what is going into and coming out of the database.

For example, consider a function that takes user data retrieved from the database and parses a created_at field to convert it from an integer into a JavaScript date object. That might look something like this:

const normalizeUser = (obj) => {
  return { ...obj, created_at: new Date(parseInt(obj.created_at)) };
};

The list goes on and on, but that could be a lot to manage as your application grows.

Or Use an ORM!

As your application grows in size and complexity, it likely will make sense to make use of some other library that can handle this typecasting logic for you. These libraries are called ORMs, which stands for object-relational mapping. That's a fancy term for a tool that makes it easier to move from database to database without needing to change the syntax used to manipulate data in the database.

At Grouparoo, we use Sequelize to interact with our application database.

Here's a simple example that does essentially what we were doing in the introductory example using Sequelize:

const faker = require("faker");
const path = require("path");

const { Sequelize, Model, DataTypes } = require("sequelize");

const sequelize = new Sequelize({
  dialect: "sqlite",
  storage: path.join(__dirname, "mydb.sqlite"),
});

const userAttrs = {
  email: faker.internet.email(),
  firstName: faker.name.firstName(),
  lastName: faker.name.lastName(),
};

class User extends Model {}

User.init(
  {
    email: {
      type: DataTypes.STRING,
      allowNull: false,
    },
    firstName: DataTypes.STRING,
    lastName: DataTypes.STRING,
  },
  { sequelize, modelName: "user" }
);

const run = async () => {
  await sequelize.sync();
  const users = await User.findAll();
  users.map((user) =>
    Object.entries(user.dataValues).map(([key, value]) =>
      console.log(`${key} (${typeof value}): ${value}`)
    )
  );
};

run();

Notice here that I'm more explicit about the column types. Sequelize then handles the typecasting for me. For example, running this once, this was the object returned to me from the database:

id (number): 1
email (string): Erling_Friesen50@gmail.com
firstName (string): Easton
lastName (string): Kub
createdAt (object): Tue Apr 20 2021 13:50:17 GMT-0400 (Eastern Daylight Time)
updatedAt (object): Tue Apr 20 2021 13:50:17 GMT-0400 (Eastern Daylight Time)

Notice that it actually sent date objects for createdAt and updatedAt. (Also notice that Sequelize handled setting those values when I created the record. I didn't have to do anything.)

There are plenty of other ORMs out there. Sequelize is among the most popular for Node. Another I came across recently is Prisma.


That's all for this exploration through how types work with SQLite when using a Node.js library. I hope you learned something!


Stay up to date

We will let you know about our launch and new content.

Share this post

Twitter Logo





Get Started with Grouparoo

Start syncing your data with Grouparoo Cloud

Start Free Trial

Or download and try our open source Community edition.